The ‘BGL’ mutant (P1), characterized by its sterility and bright green leaves, was utilized as one parent, whereas the ‘WT’ parent (P2) exhibited fertility and a standard leaf phenotype ( Figure 1 ). These two parental lines were crossed to produce the F₁ hybrid generation (BGL×WT). Subsequently, Fifty F₁ plants of bright green leaf type were used to obtain BC₁F₁ (F₁×WT) generation. Leaf phenotypes of the F₁ and BC₁F₁ generations were observed and recorded at the 7-leaf seedling stage. The bright green leaf bulk (Bulk-A) and the ordinary leaf bulk (Bulk-B) were selected from the BC₁F₁ generation based on leaf phenotype. An equal quantity of young leaves was promptly harvested, immediately placed in liquid nitrogen, and stored at -80°C for subsequent RNA extraction. The experimental materials were cultivated in a basic greenhouse at Gannan Normal University (N25°47′, E114°52′).

Total RNA was extracted utilizing the Trizol Reagent (Invitrogen Life Technologies) and subsequently quantified using both the NanoDrop spectrophotometer (Thermo Scientific) and the Qubit 4.0 fluorometer. Quality control was conducted employing the Agilent 2100/4200 system (Agilent Technologies). Equivalent amounts of purified RNA from each of the 30 samples in Bulk-A and Bulk-B were pooled for RNA library construction. The libraries were constructed following the manufacturer’s protocols, and sequencing was carried out on the HiSeq 4000 platform. Raw sequencing reads were subjected to quality control using Fastp software (version 0.19.7). Clean reads were subsequently aligned to the B. oleracea reference genome (Braol_HDEM_V1.0) using BWA software. The BoBGL locus was mapped using MMAPPR software (Hill et al., 2013), briefly, Bam files were processed with the mpileup tool from the SAMtools package to generate a pileup file, which organizes the data into a position-based format displaying sequenced bases at each location. Reads were filtered based on specified minimum base and mapping quality thresholds, and allele frequencies were calculated. The resulting data were then analyzed in R for signal processing and peak identification. R plots the Loess fits and lists SNPs enriched in the mutant pool (allele frequency >0.75, Euclidean distance >0.5). These SNPs are filtered for nonsynonymous variants using Alleler and gene annotations, with results exported to a file. MMAPPR uses these optimized default values. The association analysis was conducted using the Δ(SNP-index) method with the identified SNPs and InDels.

Mature leaves were utilized for the quantification of wax content. Leaves of appropriate size were immersed in 15 mL of chloroform for 30 seconds to extract the total epidermal wax mixture. Subsequently, 20 μg of n-tetracosane (C24) was added as an internal standard. The extract was then dried under a nitrogen (N₂) stream to a volume of 1 mL and transferred to a pre-weighed and recorded gas chromatography (GC) vial. Following the evaporation of the remaining liquid under nitrogen, 100 μL of trimethylsilyl reagent was added, and the sample was derivatized in a 70°C incubator for 60 minutes. All wax samples were converted into trimethylsilyl derivatives, subsequently dried using nitrogen gas, and then weighed. The total quantity of cuticular wax was normalized to unit leaf surface area. The experiment was conducted with three biological replicates.

Fresh mature leaf tissue was collected and fixed using a 2.5% glutaraldehyde solution for 2-4 hours. Subsequently, the samples were rinsed three times with 0.1 M phosphate buffer (pH 7.0) and then treated with a 1% glutaric acid solution in 0.1 M phosphate buffer (pH 7.0) for 1-3 hours. This was followed by three additional rinses with 0.1 M phosphate buffer (pH 7.0). The samples were then subjected to a dehydration series using ethanol concentrations of 50%, 70%, 80%, 90%, 95%, and 100% (twice) for 15 minutes each. After dehydration, the samples were soaked in a 1:1 mixture of 100% ethanol and isoamyl acetate for 30 minutes, followed by an overnight soak in pure isoamyl acetate. The samples were subsequently dried and transferred to a preparative chamber for vacuum coating. The samples were photographed using a scanning electron microscope (SEM) system (FEI Quanta 450).

Genomic DNA was extracted from young leaves using the CTAB method and diluted to 50 ng/µL with 1× TE solution. The InDels in the candidate interval from BSR-seq were used to design markers to narrow down the BoBGL locus. Primers were designed utilizing Primer3 software, the sequences corresponding to the candidate intervals were extracted, with the product size specified to range from 180 to 350 bp, allowing for one mismatched base. Primers were synthesized by Wuhan Tianyi Huayu Gene Technology Co., Ltd (Wuhan, China). The 10 µL PCR mixture included 2 µL of template DNA (50 ng/µL), 5 µL of 2× FineTaq^® PCR SuperMix, 0.5 µL of each primer (10 µM), and 2 µL of sterilized ddH₂O. The PCR program was: 94°C for 5min; 9 cycles of 94°C for 30s, 60°C for 30s (decreasing by 0.5°C per cycle), 72°C for 30s; 30 cycles of 94°C for 30s, 55°C for 30s, 72°C for 30s; and a final elongation at 72°C for 10min. PCR products were then separated on 6% polyacrylamide gel electrophoresis (PAGE).

The DNA of 940 BC₁F₁ recessive plants was subjected to amplification using polymorphic primers, followed by the screening of recombinant plants. Subsequently, the genetic distance between the markers was determined, and a genetic map was constructed. Two markers proximal to the target site were identified. A physical map was then developed by utilizing the physical locations of these two markers on the B. oleracea reference genome (Braol_HDEM_V1.0) to delineate the candidate interval. Genes in the target intervals were annotated, and key ones were analyzed for expression via qRT-PCR. Specific primers for the BoBGL gene were used to amplify sequences from two parental lines, BGL and WT, using Primer5. The PCR products were cloned into the pMDTM19-T vector and transformed into DH5α cells. The candidate gene was then sequenced using the Sanger method. Subsequently, DNA sequencing reactions were conducted by Wuhan Tianyi Huayu Gene Technology Co., Ltd (Wuhan, China). The primer sequences utilized for the molecular markers in this study are detailed in Supplementary Table S1 .

P1 is a plant with bright green leaves ( Figure 1A ), and P2 is a plant with ordinary leaves ( Figure 1B ), and the wax content is significantly reduced in P2 ( Figure 1F ). The leaf surface of the hybrid F₁ plants is characterized by bright green leaves. For the BC₁F₁ (F₁×WT) offspring, the number of ordinary leaves and bright green leaves is 1:1. Furthermore, the number of common leaves and bright green leaves in BC₂F₁ was also 1:1 ( Table 1 ). It can be judged that the bright green leaf trait of cabbage is completely apparent inheritance controlled by a pair of alleles.

The BoBGL locus was mapped using a mutation mapping pipeline for pooled RNA-seq (Hill et al., 2013). A total of 122,472,880 and 223,777,954 clean reads produced from the Bulk-A and Bulk-B RNA samples were mapped to the B. oleracea reference genome (Braol_HDEM_V1.0/, http://brassicadb.cn/#/Download/), respectively ( Supplementary Table S1 ). Over 94% of the reads were mapped to unigenes, resulting in the discovery of 247,090 SNPs and 68,667 InDels between the two bulks. SNP-index was calculated according to the SNP of two progeny pools obtained by sequencing, which was used to map the gene of bright green leaf trait in Chinese kale. The Δ(SNP-index) of the two mixed pools was analyzed. The results showed that when the confidence level was 99%, the three regions C5, C7 and C8 distributed on the chromosome showed extremely significant peaks above the critical value level ( Figure 2 ). The significant correlation intervals are: C5 (24.51-26.47Mb, ~1.96Mb), C7 (1.45-2.45Mb, ~1.00Mb), and C8 (49.36-51.42Mb, ~2.06Mb) ( Table 2 ).

Subsequently, we used the molecular marker method to further verify the candidate interval. Firstly, we examined the phenotypes (bright green leaves and ordinary leaves) of a BC₁ segregating population containing 186 individuals. Secondly, we developed some InDel markers covering all candidate intervals. Genotyping of the 186 BC₁ segregating plants was conducted with 9 pairs of polymorphic InDel markers ( Supplementary Table S2 ), resulting in the creation of a genetic linkage map. The findings indicated that the gene responsible for the bright green leaf gene, BoGBL, was situated on chromosome C8 between markers INDEL109 and INDEL104 ( Supplementary Figure S1 ), with positions on the reference genome at 49.43 Mb and 50.92 Mb, as shown in Figure 3A . Therefore, the bright green leaf gene BoGBL can be preliminarily located in about 1.5 Mb interval between 49.43Mb and 50.92Mb on chromosome C8 of B.oleracea.

In order to further narrow the candidate interval, combined with the results of preliminary mapping, 940 BC1 populations (all ordinary leaf type) were screened for recombinant plants based on flanking markers (INDEL109, INDEL104), and a total of 81 recombinant plants were obtained. At the same time, four polymorphic InDel markers were used to genotype 81 recombinant plants and finely map the bright green leaf gene BoBGL in B.oleracea. The BoBGL locus was more precisely identified to be within a 102-kb region on chromosome C8 of the B. oleracea reference genome, located between the markers INDEL150 (50 787 908bp) and INDEL78 (50 890 279bp) ( Figure 3B ; Supplementary Figure S1 ).

According to the location of the BoBGL locus, there were 24 genes within these 102-kb region in the reference B. oleracea genome ( Figure 3C ). To identify the key candidates, we conducted gene annotation on 24 genes contained in this region. The results showed that among the 24 genes, only the candidate gene BolC8t52930H, named as BoCER1.C8, was related to the biosynthesis of plant cuticle wax ( Table 3 ). The gene is highly homologous to the Arabidopsis wax synthesis-related gene AtCER1 ( Table 3 ). In addition, qRT-PCR showed the gene BolC8t52930H was significantly differentially expressed in the two parents ( Figure 4 ). In summary, the gene BolC8t52930H was identified as a key candidate gene for BoBGL.

To confirm our speculation, we cloned the genomic sequence and coding sequence of BolC8t52930H from two parental plants. The findings indicated that the coding sequences in the parental organisms were completely consistent, so as the protein sequence, despite minor variations in their genomic sequences ( Figure 5A ; Supplementary Figure S2 ). We also cloned the promoter sequence of BolC8t52930H from about 3K before the onset codon from both parents. There are some differences between the two promoters in the range of -280bp to -1500bp, with the largest being -282bp to -325bp ( Figure 5A ). The interval of about 40bp is significantly different from the insertion and deletion, and the differences in other intervals are mainly SNPs (about 30 ones), and 4 gaps ( Figure 5A ; Supplementary Figure S2 ). In summary, there are significant sequence differences in promoter regions between the two parents, and therefore, we speculate that variations in the promoter region of BolC8t52930H may be responsible for the changes in gene expression in bright green leaf material.

Next, we developed an polymorphic InDel marker using the sequence differences in the interval (-282bp to -325bp) with the largest difference in gene regions between parents to analyze whether BoBGL is co-segregated with bright green leaf traits. As we expected, the co-segregation analysis of 184 BC₁ plants showed that their genotypes were completely co-segregated with their phenotypes, that is, bright green leaf individuals can amplify a 184bp fragment, while ordinary leaf individuals cannot ( Figure 5B ). Therefore, we developed a useful functional marker that can be used for marker-assisted selection of bright green leaf traits in B. oleracea.